-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Jan Hudec wrote: | On Mon, Jan 31, 2005 at 01:58:31 +0000, Jon Masters wrote: | |>On Fri, 21 Jan 2005 12:59:59 +0100, Arjan van de Ven |><arjan@xxxxxxxxxxxxx> wrote: |> |>>On Thu, 2005-01-20 at 12:54 -0600, Timur Tabi wrote: |>> |>>>I recently discovered in a driver I'm working on that there are several |>>>places where a user-space virtual address is being dereferenced |>>>successfully. I was under the impression that that should not work - |>>>user pointers should not be usuable in kernel space. |>> |>>it may work. you may be lucky. The page also may be swapped out and you |>>crash and burn |> |>>as for fc2/3; those kernels have a separate address space for user and |>>kernel, and as a result such rogue accesses will almost always crash, |>>not just sometimes. |> |>You know, I had wondered what use this patchset had other than for |>folks with 32bit systems straining under userspace limits - this is |>actually a pretty cool driver diagnostic tool in and of itself :-)
| Wait a second... Separate address spaces for user and kerenl space?
Yes. But you "need" it is you're $large_customer and you want RH to offer a solution to the lack of 32 bit userspace memory on Linux.
| That means switching page tables on each kernel entry/exit.
Yup. It's nasty. But then, some others have to do similarly nasty things - - e.g. processors with virtually indexed caches have to do a full flush (see recent Linux Devices article comparing the performance of ARM9 with and without the MMU).
| That means tlb flush on each interrupt. That would mean quite a slowdown.
They reckon it has an acceptable overhead. I guess it depends upon what you're doing with it.
Jon. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (GNU/Linux) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
iD8DBQFB/gBleTyyexZHHxERAtckAJ9eggCN2DTnkOlpDcca0eIGl3rOTACfQh1I 9NVr8GuU27duK/esuaYgGsE= =a6uS -----END PGP SIGNATURE-----
-- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
